HubSpot Live · Published 2026-09-25

The Next 5 Years Will Not Reward the Smartest People | UNBOUND 2026 | HubSpot Live

privateReview: unreviewedConfidence: high
consistencyaccountabilityaibraverynetworkingmulti-agentevaluation-harnesspromptingpersonasschedulingreliabilitycontext-grounding
Open on YouTube ↗

Summary

Overview

  • Speaker: Jay Schwedelson
  • Channel: HubSpot Live
  • Main topic: How consistency, accountability, and action outperform pure intelligence in the next 5 years.
  • Purpose: To inspire professionals to stop waiting, hold themselves accountable, be consistent, and leverage AI differently to achieve success. Jay Schwedelson argues that in a world where AI and information are easily accessible, the next five years will not reward the smartest people, but rather those who are consistent, accountable, and take action. He shares personal anecdotes about his own college struggles, overcoming negative self-talk, and leveraging AI in unconventional ways, emphasizing that accountability and doing things consistently are the keys to career and personal success.

Topic Map

The Myth of Intelligence

  • Explanation: Intelligence and academic credentials (like GPAs or degrees) don't matter as much as people think because information is widely available and AI can handle complex reasoning.
  • Key claims:
    • The next five years will not reward the smartest people.
    • Degrees and GPAs don't matter in the real world.
    • AI has already won the intelligence contest.
  • Examples:
    • Jay's college transcript showing multiple attempts at statistics and a C in advertising strategy.
  • Terminology:
    • GPA
    • AI
    • Credentials
  • Why it matters: It shifts focus away from feeling inadequate due to lack of traditional credentials toward taking action.

Accountability and Learning by Doing

  • Explanation: Learning happens by doing, and accountability forces learning to be non-optional. You don't need motivation; you need something due.
  • Key claims:
    • Accountability makes learning non-optional.
    • You don't need more motivation; you need something due.
    • People learn by doing, not just by doing their job.
  • Examples:
    • Running a podcast or a weekly newsletter to force weekly learning.
  • Terminology:
    • Accountability
    • Motivation
    • Homework
  • Why it matters: Without forced deadlines and accountability, professionals repeat the same tasks without actual growth.

Consistency Beats Intelligence

  • Explanation: Being consistent beats 95% of people because the smartest people often stop when things get hard or busy.
  • Key claims:
    • You will beat out 95% of people just by being consistent.
    • Consistency is the secret to winning.
  • Examples:
    • Podcast drop-off rates: 47% don't make it past episode 12.
  • Terminology:
    • Consistency
    • Dropout rate
  • Why it matters: Most people stop after a few weeks; consistency requires simply not stopping.

Ten Seconds of Bravery

  • Explanation: You don't need to be fearless forever; you just need to borrow bravery for ten seconds to take action.
  • Key claims:
    • You don't have to be fearless forever; you just need to push for ten seconds.
    • You need to temporarily borrow bravery.
  • Examples:
    • Ten seconds to hit send, introduce yourself, walk across a room, or post the thing.
  • Terminology:
    • Bravery
    • Fearlessness
  • Why it matters: Overcoming anxiety and taking small bold actions changes career trajectories.

Using AI Differently

  • Explanation: AI can be used in unconventional, highly specific ways to drive business and career growth, such as analyzing LinkedIn databases or creating a fictional board of directors.
  • Key claims:
    • We need to use AI differently.
    • Downloading LinkedIn databases and prompting AI to find underutilized connections.
    • Creating a board of directors of historical figures powered by AI.
  • Examples:
    • Asking AI to create a board of directors modeled after Steve Jobs, Jeff Bezos, Mark Cuban, etc.
  • Terminology:
    • AI prompts
    • LinkedIn database
  • Why it matters: Simple, 5-minute AI prompts can drastically accelerate professional networking and decision-making.

The Danger of Negative Self-Talk and Four Toxic Words

  • Explanation: Four tiny words—still, only, should, but—do a ridiculous amount of damage to a fantastic life.
  • Key claims:
    • Still, only, should, but are the worst words in human language.
    • Nobody feels like they're enough.
  • Examples:
    • 'I still make less money...', 'I only got promoted because...', 'I should be happier...', 'I am... but...'
  • Terminology:
    • Negative self-talk
    • Imposter syndrome
  • Why it matters: Recognizing and eliminating these words stops self-doubt and allows people to realize they are enough.

Key Points

AI won the intelligence contest

  • Explanation: Intelligence alone is no longer a competitive advantage because AI and information are universally accessible.
  • Evidence: AI can process vast amounts of data and answer complex queries instantly.
  • Practical implication: Focus on execution, consistency, and human accountability rather than accumulating credentials.

Accountability creates growth

  • Explanation: Forcing yourself to produce regular outputs (newsletters, podcasts, team updates) creates forced learning.
  • Evidence: Jay maintains a weekly podcast specifically because it forces him to research and learn new things every week.
  • Practical implication: Establish weekly public or team commitments that force you to learn and adapt.

Consistency beats talent

  • Explanation: Most people quit initiatives after a few weeks. Sticking with something long-term puts you ahead of 95% of people.
  • Evidence: Statistics show massive drop-off rates in podcasts, budgeting apps, and journaling within weeks or months.
  • Practical implication: Commit to long-term consistency instead of seeking instant gratification.

Ten seconds of bravery is enough

  • Explanation: Fear holds people back, but courage only needs to be sustained for brief moments.
  • Evidence: Actions like hitting send or introducing yourself only take ten seconds of courage.
  • Practical implication: Break intimidating tasks down into ten-second micro-actions of bravery.

Frameworks, Models & Processes

Weekly Things (Accountability Framework)

  • How it works: Assigning regular weekly tasks to yourself and your team to ensure continuous learning and sharing.
  • Components:
    • 5 minutes: Share 1 industry article in team Slack
    • 10 minutes: Bring 1 competitor move to team meeting
    • 20 minutes: Send boss 3 things learned this week
    • 30 minutes: Ask to email entire department 3 things happening in industry
    • 45 minutes: Test 1 new tool, prompt, tactic
    • 60 minutes: 10 minutes of weekly department meeting sharing changes
    • The Big One: Publish something every single week
  • When to use: To maintain team alignment, continuous learning, and personal accountability.

AI Board of Directors Prompting

  • How it works: Feed an AI system the public philosophies and thinking styles of 6 historical or business icons, then pressure-test personal or business decisions against them.
  • Components:
    • Perfectionist (Steve Jobs)
    • Operator (Jeff Bezos)
    • Contrarian (Mark Cuban)
    • Brand Builder (Emma Grede)
    • Provocateur (Rory Sutherland)
    • Founder (Sara Blakely)
  • When to use: When making major personal, career, or business decisions that require diverse perspectives.

Examples & Case Studies

Jay struggled through college statistics three times, earning a C+, but went on to run an advertising agency.

  • Illustrates: Academic intelligence and grades do not dictate professional success.
  • Lesson: Your beginning doesn't get a vote on your ending.

Jay downloaded his LinkedIn connections database and prompted AI to find the 25 most underutilized connections.

  • Illustrates: Leveraging simple AI workflows on existing digital assets to drive career and business opportunities.
  • Lesson: Use AI to unlock dormant relationships and hidden value in existing tools.

Jay tried Wegovy (weight loss medication) expecting effortless results, ate everything in sight, and gained weight.

  • Illustrates: Tools and shortcuts don't replace active effort and trying.
  • Lesson: You have to do the work; there are no magic shortcuts.

Actionable Takeaways

  • Immediate:
    • Download your LinkedIn data archive and run an AI prompt to find forgotten connections.
    • Establish a weekly publishing habit (newsletter, podcast, or social post) to hold yourself accountable.
    • Eliminate the four toxic words: still, only, should, but.
  • Strategic:
    • Prioritize consistency over brilliance in your career and business strategy.
    • Build artificial accountability structures into your routine so learning is non-optional.
    • Use AI as an active sounding board and advisory board rather than just a search engine.
  • Questions to investigate:
    • What weekly accountability mechanism can I set up right now?
    • Which dormant contacts in my network am I underutilizing?
    • How can I use AI to pressure-test my business decisions?

Claims Worth Verifying

  • 47% of podcasts don't make it past episode 12. (statistical)
  • 65% abandon budgeting apps within 90 days. (statistical)
  • 87% of people quit journaling within 6 months. (statistical)

Notable Quotes

"The next five years will not reward the smartest people." (at 0:00) "Accountability makes learning non-optional. You don't need more motivation. You need something due." (at 3:35) "You will beat out 95% of people just by being consistent." (at 12:22) "You don't have to be fearless forever; you just need to push for ten seconds." (at 25:01) "Your beginning doesn't get a vote on your ending." (at 45:03)

Compressed Summary

  • AI has won the intelligence race, making raw smarts less valuable.
  • Consistency and accountability outperform intelligence every time.
  • Use AI for unconventional workflows like LinkedIn database analysis and advisory boards.
  • Eliminate the four toxic words: still, only, should, but.
  • Ten seconds of bravery is all you need to change your trajectory.
  • Keywords: consistency, accountability, ai, bravery, networking
  • Core insight: In the next five years, consistency, accountability, and action will vastly outperform raw intelligence and credentials.

Core insights

5
Mechanismmedium noveltyweak evidence

A decision can be pressure-tested by prompting a model with fixed, differentiated expert personas (perfectionist, operator, contrarian, brand builder, provocateur, founder) that each supply the public philosophies and thinking style of a named figure, producing a panel rather than a single answer.

Why it matters

This is a concrete recipe for role-differentiated critique: the diversity lives in the persona specification layer, not in the model, so it can be implemented as an orchestration pattern over one base model and reused for red-teaming plans, prompts, or architecture decisions.

Generalization

Persona-differentiated panels are a cheap substitute for heterogeneous models when building critique, review, or red-team stages; the value depends on whether the disagreement is substantive or merely stylistic.

Creating a board of directors of historical figures powered by AI.
Open source video
Feed an AI system the public philosophies and thinking styles of 6 historical or business icons, then pressure-test personal or business decisions against them.
Open source video
Practicemedium noveltymoderate evidence

Recurring, externally visible deliverables (publish weekly, weekly team sharing slot, weekly tool/prompt test) function as a forcing function that makes learning non-optional, rather than relying on motivation or intent.

Why it matters

Translates directly into scheduling layers for agentic systems: a cron-driven recurring job with a committed output artifact forces continuous ingestion, retrieval, and adaptation; without a due artifact the system (or team) repeats existing behavior indefinitely.

Generalization

Improvement in long-running AI-assisted workflows may come more from a commitment/scheduling layer than from a better model; the cadence itself is the mechanism, and it should be instrumented and measured as such.

Accountability makes learning non-optional.
Open source video
You don't need more motivation; you need something due.
Open source video
Without forced deadlines and accountability, professionals repeat the same tasks without actual growth.
Open source video
Empirical Resultmedium noveltyweak evidence

Dropout is the dominant base rate: a cited 47% of podcasts do not survive past episode 12, and consistency alone is claimed to beat ~95% of people because the most capable participants quit when things get hard or busy.

Why it matters

If attrition is the binding constraint, then persistence and restart behavior — not peak capability — are the metrics worth instrumenting for long-horizon agents, and benchmarks should report survival-to-N rather than single-shot success.

Generalization

Long-running autonomous processes should be evaluated with survival curves (runs completed before abandonment/failure) in addition to per-step accuracy, mirroring human project attrition statistics.

47% don't make it past episode 12.
Open source video
You will beat out 95% of people just by being consistent.
Open source video
Mechanismmedium noveltyweak evidence

AI is used against unconventional, high-specificity data sources — exporting a LinkedIn database and prompting the model to surface underutilized connections — rather than against generic web knowledge.

Why it matters

Grounding on an exported personal/relational dataset changes the retrieval surface from public corpora to a private graph, which is where the useful signal (dormant/underused edges) lives; it also reframes the model as an analyst over a supplied dataset rather than a search entry point.

Generalization

High-value AI use cases often come from pairing the model with an exportable proprietary graph or dataset that has no default retrieval path; the engineering work is in the export, normalization, and privacy boundary, not the prompt.

Downloading LinkedIn databases and prompting AI to find underutilized connections.
Open source video
We need to use AI differently.
Open source video
Simple, 5-minute AI prompts can drastically accelerate professional networking and decision-making.
Open source video
Predictionlow noveltyweak evidence

The speaker claims intelligence is no longer a competitive moat because AI and information are universally accessible, and asserts that what remains scarce is execution, accountability, and persistence.

Why it matters

If capability is commoditized, differentiation shifts into harness-level concerns — scheduling, commitment tracking, output pipelines, and human approval steps — rather than model selection or benchmark scores.

Generalization

As base capability spreads, competitive advantage migrates to the orchestration and operating layer around the model; architectural investment should follow the scarce resource.

The next five years will not reward the smartest people.
Open source video
AI has already won the intelligence contest.
Open source video

Deep dives

4

Persona-panel critique vs single-pass decision review

Research question

Do fixed persona panels produce measurably better decisions than a single strong-model pass, or do they mainly add token cost and stylistic variation?

Why

If persona diversity is substantive, critique orchestration becomes a versioned config layer with measurable disagreement; if not, multi-agent review is expensive theater and should be collapsed to one strong pass.

Creating a board of directors of historical figures powered by AI.
Open source video
Feed an AI system the public philosophies and thinking styles of 6 historical or business icons, then pressure-test personal or business decisions against them.
Open source video
Source video

Survival-to-N curves for long-running scheduled agents

Research question

Is the reported attrition curve (47% of podcasts failing before episode 12) a reliable base rate that transfers to long-running agent runs?

Why

If attrition is the binding constraint, reliability benchmarks should report survival-to-N and restart behavior, not only per-step accuracy or single-shot success.

47% don't make it past episode 12.
Open source video
You will beat out 95% of people just by being consistent.
Open source video
Source video

Forcing-function cadence and artifact feedback

Research question

Does a mandatory recurring-output cadence actually change outcomes, or does it just produce low-value artifacts on schedule?

Why

The cadence is only a mechanism if artifacts feed back into later runs; otherwise scheduling becomes ritual compliance and observability should measure reuse, not just completion.

Accountability makes learning non-optional.
Open source video
You don't need more motivation; you need something due.
Open source video
Without forced deadlines and accountability, professionals repeat the same tasks without actual growth.
Open source video
Source video

Privacy-preserving grounding over exported relational graphs

Research question

What is the minimum data-access and privacy boundary needed to safely run analysis over exported relational graphs?

Why

Private-graph use cases move engineering effort into export, normalization, redaction, and retention; without a boundary, the most useful grounding surfaces are also the most compliance-risky.

Downloading LinkedIn databases and prompting AI to find underutilized connections.
Open source video
We need to use AI differently.
Open source video
Source video

Article ideas

4

Your AI Review Panel Is Probably Expensive Groupthink

Multi-persona review panels improve decisions only when the personas disagree on decision-relevant criteria; otherwise they add latency, cost, and a false sense of rigorous evaluation.

Angle

Measure disagreement entropy and substantive-objection rate across a single-pass baseline and an N-persona panel on the same decision set.

Source video

Consistency Is an Infrastructure Requirement, Not a Motivational Trait

The dropout statistic is not a lesson in willpower but a signal that long-running AI workflows need a scheduling and commitment layer that survives interruption, restart, and operator absence.

Angle

Translate podcast attrition into survival-to-N engineering metrics for scheduled agent jobs.

Source video

Grounding on Private Graphs: The Prompt Is the Easy Part

The valuable AI use cases over exported relational databases are gated by export, normalization, consent, and retention design, not by prompt cleverness.

Angle

Break down the pipeline from raw LinkedIn-style export to redacted prompt context and identify where projects fail.

Source video

Accountability Is a Due Artifact, Not a Personality Trait

Forcing functions work because they create an external due artifact; agentic systems should implement deadlines and deliverables as first-class orchestration objects rather than relying on user motivation.

Angle

Convert 'you need something due' into cron-driven agent job contracts with observable artifact survival.

Source video

Project ideas

4

Persona Panel A/B Harness

beyond-evals

A fixed N-persona panel will surface decision-relevant objections that a single strong-model pass misses, measured by blind judge revision rate and substantive-objection classification, without collapsing into stylistic variation.

Proof of concept

Run the same decision memo through one baseline prompt and a six-persona panel (perfectionist, operator, contrarian, brand builder, provocateur, founder); classify each objection as substantive or stylistic and have blind raters select the more actionable critique.

Measurement

Decision revision rate, substantive-objection count per dollar, blind preference, and persona agreement entropy across 50 decisions.

Source video

Survival-to-N Agent Reliability Metric

beyond-evals

Scheduled agent jobs exhibit a survival curve where most failures or abandonments happen before N cycles, and survival-to-N predicts downstream utility better than per-step accuracy.

Proof of concept

Instrument a repeated weekly research or summarization job; log cycle completions, aborts, and artifact quality; plot survival curves for 100 scheduled runs.

Measurement

Median cycles survived, survival-to-10 rate, and correlation between survival and cumulative artifact reuse.

Source video

Forcing-Function vs On-Demand Knowledge Artifacts

movement-lab

A mandatory recurring-output workflow produces more cumulative reusable knowledge artifacts than an on-demand workflow over the same task period, even when total token budget is equal.

Proof of concept

Compare two agent workflows: one must produce a weekly synthesis artifact, one runs only when prompted; both start from the same corpus and token budget.

Measurement

Artifact count, retrieval reuse count, downstream task success on held-out queries, and operator-perceived value.

Source video

Redacted Graph Grounding Pipeline

gatehouse

Field-level redaction and retention limits on an exported relational graph preserve most underutilized-connection signal while reducing privacy exposure.

Proof of concept

Prototype a local pipeline that ingests a synthetic LinkedIn-like connection export, applies field-level redaction, and prompts a model to surface dormant connections; compare against an unredacted baseline.

Measurement

Precision and recall of surfaced dormant connections, sensitive-field leakage count, and latency overhead.

Source video

Architectural implications

2

The speaker describes assembling a panel of named archetype personas (perfectionist, operator, contrarian, brand builder, provocateur, founder) to review decisions.

Before

A single model response is treated as the evaluation signal.

After

Evaluation responsibility moves into an orchestration layer that instantiates multiple fixed persona critics over the same artifact and aggregates or contrasts their outputs.

Consequence

Persona definitions, ordering, and aggregation become first-class, versioned configuration; cost and latency scale with panel size, so the trade-off between critique diversity and budget becomes an explicit design parameter.

Source video

Continuous weekly publication and weekly team sharing are framed as the mechanism that forces ongoing learning.

Before

Improvement is attributed to model or individual capability accumulation.

After

A scheduling/commitment layer with required recurring output artifacts sits between the system and its data, forcing periodic ingestion and synthesis.

Consequence

Observability must track cadence adherence and artifact survival (how many consecutive cycles completed) rather than only per-invocation quality.

Source video

Tradeoffs and failure modes

3

Multi-persona review panels

Benefit

Multiple fixed thinking styles pressure-test a decision and surface considerations a single pass may skip.

Cost or risk

The summary provides no outcome evidence that panel diversity improves decisions; the risk is additive cost, latency, and more opinions without added signal, plus persona outputs collapsing toward agreement.

Creating a board of directors of historical figures powered by AI.
Open source video
Source video

Consistency as a strategy

Benefit

Sticking with a recurring initiative is claimed to outperform most participants who quit early.

Cost or risk

Consistency without a feedback loop rewards repetition rather than progress; the summary itself warns that absent forced deadlines people repeat the same tasks without actual growth.

Without forced deadlines and accountability, professionals repeat the same tasks without actual growth.
Open source video
Source video

Grounding on exported personal datasets

Benefit

Querying a proprietary relational graph surfaces underutilized connections that generic retrieval would not.

Cost or risk

The summary is silent on consent, data-retention, and compliance handling for exported third-party data such as a LinkedIn database.

Downloading LinkedIn databases and prompting AI to find underutilized connections.
Open source video
Source video

Open questions

4

Do fixed persona panels produce measurably better decisions than a single strong-model pass, or do they mainly add token cost and stylistic variation?

Why unresolved

The summary describes the prompt construction but reports no evaluation, comparison, or outcome data.

Research direction

Run an A/B harness on the same decision set: single-pass baseline vs. N-persona panel, scored by blind human raters or a held-out judge, measuring disagreement rate, decision revision rate, and cost per decision.

Source video

Is the reported attrition curve (47% of podcasts failing before episode 12) a reliable base rate that transfers to long-running agent runs?

Why unresolved

The summary cites the statistic without a source and does not connect it to AI systems.

Research direction

Instrument repeated scheduled agent jobs and publish survival curves (fraction of runs surviving to N cycles) to compare against the human-project dropout baseline.

Source video

Does a mandatory recurring-output cadence actually change outcomes, or does it just produce low-value artifacts on schedule?

Why unresolved

The cadence framework is asserted as effective; the summary offers no measurement of learning or performance gain attributable to it.

Research direction

Compare a scheduled forcing-function workflow against an identical on-demand workflow on the same task, measuring cumulative knowledge artifacts, retrieval reuse, and downstream task success.

Source video

What is the minimum data-access and privacy boundary needed to safely run analysis over exported relational graphs?

Why unresolved

The LinkedIn database example is stated as a use case with no treatment of consent, redaction, or retention.

Research direction

Prototype a local grounding pipeline with field-level redaction and retention limits, then test whether scrubbing sensitive attributes degrades the connection-surfacing results.

Source video

Key claims

8
predictionVerification needed

The next five years will not reward the smartest people.

Evidence

The next five years will not reward the smartest people.

Question

What measurable proxy for 'reward' (income, promotion rate, venture outcomes) would falsify this over a five-year window?

Source video
causalVerification needed

AI and information availability have removed raw intelligence as a competitive differentiator.

Evidence

AI has already won the intelligence contest.

Question

On which cognitive benchmarks does AI match or exceed professionals, and does that parity actually erode the economic return to measured intelligence?

Source video
factualVerification needed

47% of podcasts do not make it past episode 12.

Evidence

47% don't make it past episode 12.

Question

What is the cited source and sample for the 47% episode-12 survivorship figure?

Source video
causalVerification needed

Accountability, not motivation, is what makes learning happen.

Evidence

Accountability makes learning non-optional.

Question

Does an externally imposed deadline produce more skill or output gain than intrinsic motivation alone under controlled comparison?

Source video
comparativeVerification needed

Consistency alone puts a person ahead of roughly 95% of others.

Evidence

You will beat out 95% of people just by being consistent.

Question

In which population and initiative is the 95% figure measured, and over what time horizon?

Source video
causalVerification needed

Very short AI prompts against specific datasets can substantially accelerate networking and decision-making.

Evidence

Simple, 5-minute AI prompts can drastically accelerate professional networking and decision-making.

Question

Compared with unassisted effort, what measurable improvement in decision quality or connection outcomes results from such prompts?

Source video
opinionVerification not requested

The words still, only, should, and but actively damage outcomes.

Evidence

Still, only, should, but are the worst words in human language.

Source video
causalVerification needed

Fear is surmountable with brief, bounded bursts of courage rather than permanent fearlessness.

Evidence

You don't have to be fearless forever; you just need to push for ten seconds.

Question

Does decomposing intimidating actions into short commitment windows increase the rate of completed high-anxiety actions?

Source video

Connections

4